Desktop GPU Performance Hierarchy Table (Archive)

[cangelini]

Here is a resource to help you judge if a graphics card is a reasonable value: The gaming GPU hierarchy chart groups GPUs by performance.

Desktop GPU Performance Hierarchy Table : Read more

 

[adamovera]

Archived comments are found here: http://www.tomshardware.com/forum/id-3164715/desktop-gpu-performance-hierarchy-table.html

 

[bit_user]

I have a few doubts about this latest iteration. Would you please give us a bit of insight into the data & methodology you're using?

Otherwise, this is a valuable resource to which I've long referred people. Thank you for maintaining it.

 

[Jamroast]

http://www.game-debate.com/news/22146/amd-slashes-almost-half-off-the-radeon-pro-duo-asking-price-now-available-for-799

 

[Computered]

This list is so wrong it is beyond words. The 480 matches the 1060 not a tire below. The Fury beats the 980 and comes close to the 980ti. Also the 480 price is way wrong it can be had for around $225 not $300+. I could go on pointing out errors but you get the point. Was this just sloppy work or an intentional move?

 

[mubin]

@COMPUTERED, i agree with you too. 1060, 480 and 390 all same tier card alone with gtx690. They are not high tier. This chart needs to correct.

 

[Number42]

As has been said. What the ???? RX 480 a tier below the 1060, that's just a lie and misleading.

 

[bit_user]

I was thinking about it, and the way I'd compile such a table is by keeping a database of all benchmark results. First, for each GPU, discard the results that fall outside the "playable" range. Then, use K-means clustering to separate them into tiers. Try this with a few different values for k, until the result segments most of the GPUs of each generation into separate tiers.

The only tricky part is that you'd have to tweak the distance function to ignore components where one or both GPUs were lacking that particular benchmark.

One nice thing about it is that the cluster centers would serve as estimates of framerates that GPUs would get in games & configurations on which they were never tested. All that's needed would be for some of the GPUs in that cluster to have results.

Finally, I think it might work better to use 1/framerate.

 

[kicsako]

Another nvidia biased chart. tomshardware is getting ridiculous

 

[Dunzaus]

Gtx 1070 should Not have it own tier above 980 ti and Titan X . So much wrong in this chart

 

[pb4udie]

You used to have the best graphics comparison chart on the web and now u come up with this? This looks like someone created a chart based on some googling instead of doing some actual testing. And even with google input only this chart looks more like "cheaper is weaker" than actual performance wise chart.

What test was used? what is the score of each card in most common used resolution?
Is it April first?

 

[ralanahm]

This is way off.

 

[werberman]

Missing the Titan Z entirely

 

[FaceBob]

Gotta agree with everyone else here, this hierarchy chart is just all kinds of wrong, particularly with the 1060/480 comparison. Based on *up-to-date* benchmarks over a broad range of titles (and they're not that hard to find), the two cards couldn't be much more equal - on performance AND price! Even the somewhat higher power consumption of the 480 isn't as big a deal as many people seem to get all hyperbolic about.

Really wish this chart can be fixed as it seems highly misleading. Whether intentional or not, it looks like serious bias is going on.

 

[Seltymar]

What's the Radeon Fury Namo ?
The link of the 480 to amazon is on "Currently unavailable" for months. Even the Gaming X is way cheaper than that.
This shows the lack of professionalism here.
Kind of desappointed by Tom's Hardware, I used it a lot to build my computer...

 

[CaptainTom]

I just stumbled upon this chart and had to comment: This is the biggest joke I have ever seen.

The Fury Nitro is consistently trading blows with the 1070 in all of the latest games, and the Fury X almost always beats it (And sometimes comes close to the 1080).

There are many more piles of inaccuracies all over the chart though.

 

[Pompompaihn]

Just beating the dead horse...I have a Haswell i5 desktop with a RX480 and a Skylake i7 1060 laptop with the same amount of RAM...they're dead even with slight differences based on games you're playing. If I hook them to the same external monitor, I use the exact same settings on both. This is just silly. Over the summer I might have agreed, but new drivers have pushed the 480 passed the 1060 for me in a few games even. Fix this. And the insane $300+ price you have for the RX480 form Amazon....

 

[AndrewJacksonZA]

Here is a resource to help you judge if a graphics card is a reasonable value: The gaming GPU hierarchy chart groups GPUs by performance.

Seriously Chris? Seriously seriously seriously? I'm a big fan of you guys, but I'm now upset with you for not fixing glaring errors in your charts, least of all obvious typos. C'mon, you guys have a higher standard than this!! 😱 I'm really ragging on you and being persistent because I see you as one of the higher quality consumer technical websites around, don't slack off in an important area that many people will use as a reference!

For fellow readers, in case you missed it, here is Hardware Canucks' updated review in early December 2016: http://www.hardwarecanucks.com/reviews/video_cards/gtx-1060-vs-rx-480-an-updated-review/

Hardware Canucks Youtube video review link: https://www.youtube.com/watch?v=gEw3CaNSbUo [video="https://www.youtube.com/watch?v=gEw3CaNSbUo"][/video]

My comment from before when the chart got updated previously still stands:

Hi guys

Thank you for keeping your chart updated, it's useful for me and I'm sure for many other people too. 🙂

I don't understand your reasoning for putting the GTX 1060 one tier above the RX 480. Please could you explain it to me? Based on benchmark evidence from your 1060 review http://www.tomshardware.com/reviews/nvidia-geforce-gtx-1060-pascal,4679-3.html as well as other reviews on the web (because you guys didn't include a Vulkan API in your benchmarks like, saaaaayyyyyy... Doom) I disagree. I would put them on the same tier.

If I had to only focus on DX11, I'd put the 1060 on a tier above the 480.
If I had to only focus on DX12 and Vulkan, I'd put the 480 on a tier above the 1060.
If I had to focus on all three, I think that the same tier is fair.

In case any readers are looking for a Doom benchmark and somehow hasn't read one yet, go to [H]ardOCP: http://www.hardocp.com/article/2016/07/19/nvidia_geforce_gtx_1060_founders_edition_review/4
I'm pointing you to that site specifically because they seem, to me, to be a site that's biased towards Nvidia, but they do have the integrity to be fair and honest when the evidence demonstrates the opposite, as it is in this case.

@Toms team regarding Doom:
On 19 July 2016 Igor said that you guys were waiting on id to release a patch to enable Vulkan async compute on Nvidia hardware.

On 23 July 2016 Jason Leavey posted that Vulkan is now supported on AMD and Nvidia hardware with the appropriate drivers - it's just that async compute isn't supported on Nvidia hardware yet but there'll be an update "soon," just as Igor said.

That was over a month ago. Has id given any evidence that the patch is coming "soon?" Any date estimate whatsoever, perhaps? Is it ever coming out? Please add the Doom Vulkan benchmarks.

 

[elw]

I think everyone can use google to see NEW benchmarks or see on youtube cards on their new drivers in games. Compare them all to get more accurate result than this. Time to stop to take Tom's charts too seriously. Kinda pitty I ussually found Tom's site quite good when comparing processors and gfxs. Well times are changing.

 

[Borisblade7]

Man so much of this is quite wrong, wtf? Usually you guys have a great list to reference. The RX480 is equal to a 1060 and beats it in games with vulcan, dx12, and other newer features it should be in the same tier, on top of that you have the price completely wrong showing some insane $330 when its more like $230 for the 8gb model, less for 4gb. And several other discrepancies. Not to mention I saw several pretty obvious spelling errors. Someone needs to check their info and proofread.

 

[Luis Melo]

Where is Radeon Pro Duo.

 

[Tonyooki]

Is this a satire? I stopped reading when I see HD7990 on par with 980TI.

 

[Cok3Zer0]

Maybe add the INTEL HD Graphics 530 (integrated) ???

 

[Clariska]

Well for us its different in prices.I can get 2x R9 295X2 for the price of 1 titan X nvidia gfx card so in general the radeons are a much better bargain for me.Good not the less(i swear i need to move to freaking US)stuff so cheap there -_-

 

[NightAntilli]

Drop all nVidia Pascal cards one tier and you have a proper tier list.